Webwork and Stripes: A Simple Performance Comparison

Well, it took a bit longer to run the load tests than i thought (due to finishing some work for the next build of the site), but we finally got around to it yesterday.

Like I mentioned in the previous post..I decided to rewrite one of our pages in Stripes, and perform a load test to compare the Stripes version to the WebWork 2.2.2 version.

To briefly recap the last post..the ONLY differences in the 2 versions of the page are the Action class and the jsp. Thats it. All other layers...Spring, iBatis, data model, etc are the EXACT same in both versions. Even the business logic within each Action class is the exact same. The only diff is the each Action has a different superclass, and Webwork is configured with xwork.xml, and Stripes uses annotations. Thus most all of the application code (non-framework) is the same. As for the jsp..the webwork version uses webwork tags exclusively, and the Stripes version uses a combination of Stripes form tags, and JSTL. Since most of the application code is the same, i believe any differences in test results should be due to the differences in Webwork and Stripes.

We ran the load test on a test server, not a real prod server. I know the final numbers wouldnt mimick live numbers, but it is still a valid as a relative comparison between the 2 versions.

I'll be honest...after reading about performance issues with OGNL, and seeing how some of the framework code looks..i was kind of expecting Stripes to be a bit faster. From working with the two..that was just my gut feeling.

However, I will admit I was pretty surprised by the results. In our tests, the Stripes version of the page was 3-4 times faster than the webwork version! Thats a pretty significant difference.

Now..i'm not trying to say the Stripes is always that much faster than webwork, or that in all cases Stripes will be faster (though i do indeed believe Stripes will usually be faster in most cases)..this test was just comparing the 2 frameworks on 1 page in our site. So make of this test what you will. I'm just reporting what we found in our case..nothing more. It is good ammo for me to use to push for officially using Stripes in our application. Since the site has only been complete for 6 weeks, its a little early to propose a big architecture change. However, i plan to convert the site over in my free time as I can...the sooner I do it, the less will have to be changed later.

Having worked with webwork since December, I feel i have a good handle on how to develop with it. I really dont have any big complaints, either, as I enjoyed using it. As good as webwork is, I just think Stripes is one step better. If you can get the benefits of the self-documenting nature of annotations (and no config files), a little speadier development, and a performance boost..I think you have to consider Stripes a serious player in your web framework decision.

I hope people continue to discover how easy Stripes is to use and the user base continues to grow. I think this framework really deserves the attention.

As I continue to migrate the site (unofficially) from webwork to stripes, i will post any problems i run across, as well as any "this is awesome" moments, too. I expect there to be a lot more of the latter.